You are viewing content from a past/completed QCon -

Presentation: Products And Prototypes With Keras

Track: Modern Learning Systems

Location: Mountbatten, 6th flr.

Day of week:

Slides: Download Slides

Level: Intermediate

Persona: Data Scientist

What You’ll Learn

  • Learn foundational knowledge and principles connected to deep learning.
  • Gain practical knowledge applying Keras (a framework used in deep learning) against common problems in the space.
  • Understand how products/offerings like Amazon’s GPU instances, Tensorflow, and Theano can help you build complex solutions with minimal investment in time and money.


In this talk Micha will show how to build a working product with Keras, a high level deep learning framework. He'll start by explaining deep learning at a conceptual level, before describing the product requirements. He'll then show code and discuss design decisions that demonstrate how to train and deploy the model. In the process, he'll place Keras in context in the deep learning framework ecosystem, that includes Tensorflow, MXNet and Theano.


What’s the focus of the work you are doing today?


At Fast Forward, we do Quarterly Reports on new technologies and AI.

So what does that mean? We’ve done things such as Natural Language generation from structured data (and how to make it fast using probabilistic data structures on massively distributed systems). We’ve done image classification and automatic text summarization using Recurrent and Convolutional Neural Networks. Most recently we have done probabilistic programming (which is a new paradigm in programming languages that add bayasen primitives). For all of these things, we write in depth reports that includes interview with the people working on the technology, prototypes, and reference implementations. We take the approach that we don’t understand the space until we can code it.

So much of my role is really looking for what’s new in AI. I try to focus on what’s possible today that maybe was wasn’t possible a year ago.


Can you tell us what you’re going to focusing on with your talk?


What I’m planning to do is go through a deep learning case study that leverages Neural Networks / Deep Learning. The thing I’d like to get across is that many developers may think these approaches are simply out of reach, but the idea is they really aren’t. We will take these modern frameworks and iterate on solutions solving them in this talk as we go. The interesting thing is that by combining modern frameworks and other things like Amazon’s GPU instances it’s really incredibly simple to leverage deep learning today.


Who are you targeting with your talk?


The audience should be people who will take an idea and then spent the weekend hacking on it. It’s not necessarily a goal to give them what they need to build a production system, but it will be enough to give them the tools to move that way (to play).

I’ll try to touch briefly on some of the mathematical intricacies (maybe 10 seconds at a time scattered throughout the talk), so that if there are ML experts in the audience you’ll still be interested. But that’s not going to be the primary focus. The focus is to jumpstart developers with deep learning.


What do you want someone with a base foundation in ML to walk away with?


I want them to have foundational knowledge to start approaching their use cases with the framework of deep learning.

Speaker: Micha Gorelick

Research Engineer @FastForwardLabs, Keras Contributor

Research engineer at Fast Forward Labs, Keras contributor. Previously at

Find Micha Gorelick at

Preliminary tracks

Discover some of the topics you will see at QCon London. *The schedule is subject to change

Event-Based Architectures: The Hard Parts

Scaling Security, from Device to Cloud

From Remote to Hybrid Teams: Return to Office?

Architectures You've Always Wondered About

Building High Performing Teams

Performance/Mechanical Sympathy

Next Generation Microservices: Building Distributed Systems the Right Way

Modern Data Pipelines & Streams

The Cloud Operating Model

Chaos and Resilience: Architecting for Success

Scaling Frontend

Modern JVM Innovations

MLOps: Implementing ML Across the Enterprise

Architecting a Modern Financial Institution

Crafting the Developer Experience